Fragonomics: fragment-based drug discovery
Edward R Zartler and Michael J Shapiro
The use of smaller molecules (fragments) in the drug discovery
process has led to success in delivering novel leads for many
different targets. This process is a highly integrated process,
starting from library design to screening and medicinal
chemistry. An overview of this process is presented with
particular emphasis placed on the NMR aspect of screening.
Addresses
Lilly Research Laboratories, Eli Lilly & Co., Indianapolis, IN 46285, USA
Corresponding author: Zartler, Edward R (zartler_edward@lilly.com)
Current Opinion in Chemical Biology 2005, 9:366–370
This review comes from a themed issue on
Next-generation therapeutics
Edited by Chris J Vlahos and Michael Coghlan
Available online 31st May 2005
1367-5931/$ – see front matter
# 2005 Elsevier Ltd. All rights reserved.
DOI 10.1016/j.cbpa.2005.05.002
Introduction
Fragment -
1
fragÁment: a part broken off, detached, or
incomplete.
Fragonomics -
1
fragÁ3Á’nO-miks: a highly integrated lead
generation approach using small, relatively simple mole-
cules.
There are two approaches that are increasingly used to hit
the ever-more challenging targets that exist in today’s
pharmaceutical portfolio: targeted libraries and fragments
[1
]. The theory behind targeted libraries is that small
libraries representing scaffolds that are known to inhibit a
given class of target have a very high probability of
delivering potent inhibitors of that target. Fragonomics
takes the opposite approach. It uses small, simple mole-
cules screened at high concentrations to find molecules
that can be developed eventually into drugs [2,3]. This
approach is a highly integrated, collaborative effort,
requiring the combined efforts of the many scientific
disciplines present in today’s pharmaceutical company:
computational chemists, structural biologists, organic
chemists, and biologists. The use of fragments in both
the lead generation and lead optimization regimes is
becoming increasingly widespread, and the impact of
fragments is hard to ignore [4,5]. Fragonomics can be
broken into three stages: library design, screening, and
medicinal chemistry (Figure 1), which are each discussed
below.
Library design
Smaller is better
The ‘Screen for Drug’ paradigm, which dominates the
pharmaceutical industry currently, is best illustrated by
the 470 lead-drug pairs studied by Hann et al.[6]. It was
found that 78% of these pairs increased in mass by an
average of 4–5 heavy atoms. Remarkably, 22% of these
pairs decreased in mass, by an average of $4 heavy atoms,
indicating that the initial libraries were too large. This
shows the highly restricted chemistry space available for
lead optimization and may be one of the reasons that the
advent of high-throughput screening (HTS) and massive
libraries have not reaped the increases in productivities
that were expected [1
]. Lead optimization often leads to
larger and more lipophilic molecules due to exploitation
of hydrophobic interactions to increase potency [7
].
However, too much hydrophobicity or molecular weight
can contribute to poor pharmacokinetic and pharmaco-
dynamic (PK/PD) properties. This was one of the moti-
vations for Lipinski’s Rule of 5 [8,9]. Fragments start with
smaller molecules, providing more available chemical
space for lead optimization. Additionally, Vieth et al.
showed that orally available drugs are, as a class, smaller
and less complex than drugs with other routes of admin-
istration [7
], so fragonomics holds even more opportu-
nity in the discovery of orally bioavailable drugs.
The complexity of fragments
Current libraries have too much functionality, decreasing
the chance of finding a lead in the first place, and then
severely limiting optimization if a lead is found [10]. In
2001, Glaxo’s libraries were more complex compared with
both marketed drugs and leads [6]. Hann et al. studied
molecular complexity and its effect on the success of
screening. The chance of finding a detectable and unique
binding mode is dependent on the chance of measuring
binding and the chance of having a unique binding mode
or match. The chance of measuring binding increases
with the complexity of the ligand. The chance of having a
unique match goes up as the molecule becomes increas-
ingly complex, but there is also a greater probability of
negative interactions. Figure 2 shows the trade off
between detecting binding and a unique match. Hann
et al. are not trying to predict the optimum complexity of a
molecule for screening; instead, their point is that if the
molecules start simpler than typical, there is a better
chance of finding both detectable binding and a unique
binding mode. Lead molecules that are simpler also give
more available chemical space for optimization, especially
Current Opinion in Chemical Biology 2005, 9:366–370 www.sciencedirect.com